[ANALYSIS] Goodrich impingement study and NASA’s LEWICE program (see section 1.16.2) also predicted a sparse, rough ice
accretion aft of the deicing boot on the lower wing surface for some of the tested conditions. However, no ice
accretion aft of the deicing boot was noticed during the natural icing certification tests. (See section 2.5.1.)
Although it is possible that some of the drag observed in the accident airplane’s performance was the result of a
sparse, rough ice accumulation aft of the deicing boot on the lower wing surface, it was not possible to positively
determine whether the accident airplane’s ice accretion extended beyond the deicing boot coverage.
145
stitchlines. During tests conducted at a TAT of 26o F, a small, but prominent (½ inch) ridge of
ice frequently appeared on the forward portion (0.5 to 1 percent MAC) of the leading edge
deicing boot’s upper surface. The IRT test results were used in NASA’s computational studies, which indicated
that these pronounced ice ridges tended to act as stall strips, creating more disrupted airflow over
the airfoil’s upper surface, further decreasing the lift produced by the airfoil, and resulting in a
lower stall AOA than the rough ice accretions alone. NASA’s computational study data
indicated that a thin, rough ice accretion with a small, prominent ice ridge can result in a lower
stall AOA and a more dramatic drop off/break than the 3-inch ram’s horn ice shape commonly
used during initial icing certification testing. The implications of this finding for FAA icing
certification criteria are discussed in section 2.6.2.
The accident airplane’s performance displayed evidence of adverse effects on
both lift and drag during the airplane’s descent to 4,000 feet msl. The degradation exhibited by
the accident airplane was consistent with a combination of thin, rough ice accumulation on the
impingement area (including both upper and lower wing leading edge surfaces), with possible
ice ridge accumulation. Thus, based on its evaluation of the weather, radar, drag information,
CVR, existing icing research data, and postaccident icing and wind tunnel test information, the
Safety Board concludes that it is likely that Comair flight 3272 gradually accumulated a thin,
rough glaze/mixed ice coverage on the leading edge deicing boot surfaces, possibly with ice
ridge formation on the leading edge upper surface, as the airplane descended from 7,000 feet msl
to 4,000 feet msl in icing conditions; further, this type of ice accretion might have been
imperceptible to the pilots.
[ANALYSIS] The IRT observers further noted that IRT lighting conditions and cloud (spray) type greatly
affected the conspicuity of the ice accumulation, making it difficult to perceive the ice
accumulation during the icing exposure periods. NASA-Lewis’s scientists described the IRT ice
accretions as mostly “glaze” ice, like mixed or clear ice in nature, although it looked slightly like
rime ice when the IRT was brightly lighted for photographic documentation of the ice accretions
because of its roughness. The Safety Board notes that it is possible that such an accumulation
would be difficult for pilots to perceive visually during flight, particularly in low light conditions. This type of accumulation would be consistent with the accident airplane’s CVR, which did not
173 Although the Safety Board considered other possible sources for the aerodynamic degradation
(such as a mechanical malfunction), the physical evidence did not support a system or structural failure, and the
FDR data indicated a gradual, steadily increasing performance degradation that was consistent with degradation
observed by the Safety Board in data from events in which icing was a known factor.
174 All pilot reports indicated moderate or less ice accretions, except the pilots of NW flight 272,
who reported that they encountered a trace of rime ice during the descent, then encountered moderate-to-severe
icing at 4,000 feet msl about 2 minutes after the accident.
175 These TATs are equivalent to SATs of 21o F (-6o C) to 25.5o F (-3o C).
144
record any crew discussion of perceived ice accumulation and/or the need to activate deicing
boots during the last 5 minutes of the accident flight. The location of rough ice coverage observed during the icing tunnel tests varied,
depending on AOA; at lower AOAs, the ice accretions extended farther aft on the upper wing
surface (to the aft edge of the deicing boot on the upper wing surface, about 7 percent of the wing
chord at the aileron midspan), whereas at higher AOAs, the ice accretions extended farther aft on
the lower wing surface. In some IRT test conditions, sparse feather-type ice accretion extended
aft of the deicing boot coverage on the lower wing surface (which extends to about 10½ percent
of the airfoil chord at the aileron midspan) as far as 30 to 35 percent of the airfoil’s chord.176
The density of the rough ice coverage also varied, depending on the exposure
time; a sparse layer of rough ice usually accreted on the entire impingement area during the first
30 seconds to 1 minute of exposure, and the layer became thicker and more dense as exposure
time increased.
[ANALYSIS] In addition, a hazardous situation may develop even if deicing boots are operated
throughout an icing encounter as a result of ice accretions on an airplane’s unprotected surfaces,
such as aft of the deicing boots. As previously noted, the B.F. Goodrich impingement study,
NASA’s LEWICE calculations, and NASA IRT tests indicated that a light accretion may occur
on the unprotected lower wing surfaces aft of the deicing boot on the EMB-120. However,
Embraer representatives stated that such an ice accretion would result in only a trace of ice
accumulating aft of the deicing boots and would have a minimal aerodynamic penalty in drag
only. Although there was no evidence of ice accretion aft of the deicing boot during the EMB120 certification natural icing tests and it was not possible to determine whether the accident
airplane’s ice accretion extended aft of the deicing boot coverage, it is possible that ice accretion
on the unprotected surface aft of the deicing boot could exacerbate a potentially hazardous icing
situation. Based on icing and wind tunnel research and information from the Westair
incident, the Safety Board concludes that it is possible that ice accretion on unprotected surfaces
and intercycle ice accretions on protected surfaces can significantly and adversely affect the
aerodynamic performance of an airplane even when leading edge deicing boots are activated and
operating normally. Thus, pilots can minimize (but not always prevent) the adverse effects of ice
accumulation on the airplane’s leading edges by activating the leading edge deicing boots at the
first sign of ice accretion. It is not clear what effect residual ice/ice accretions on unprotected
nonleading edge airframe surfaces have on flight handling characteristics. Because not enough
is known or understood about icing in general, and especially about the effects of intercycle and
residual ice, the Safety Board believes that the FAA should (with NASA and other interested
aviation organizations) conduct additional research to identify realistic ice accumulations, to
include intercycle and residual ice accumulations and ice accumulations on unprotected surfaces
aft of the deicing boots, and to determine the effects and criticality of such ice accumulations;
182 According to the pilots of Westair flight 7233, they were aware that they were operating in
“icing conditions;” they stated that they observed ice accumulating on the airplane and had activated the leading
edge deicing boots when the airplane entered the clouds during their departure.
153
further, the information developed through such research should be incorporated into aircraft
certification requirements and pilot training programs at all levels. The Safety Board considers it likely that future ice detection/protection systems
will decrease the hazards associated with icing by incorporating ice detection and protection
(automatic activation of deicing boots or anti-icing systems) for individual surfaces, including
the horizontal stabilizers, of all airplanes certificated for flight in icing conditions.
[ANALYSIS] Further, as has been recognized for 50 years or more, and demonstrated in
accidents in the 1970s, 1980s, and early 1990s, and then again in the Comair flight 3272
accident, surface roughness/ice accretions that may be imperceptible or appear insignificant to
191 The FAA has since required manufacturers of turbopropeller-driven airplanes to develop visual
cues for SLD icing; however, the cues were based on very limited testing. Thus, the Safety Board is not convinced
that such cues will exist for all icing conditions outside the appendix C icing envelope.
168
pilots can adversely affect the operation of the airplane. However, because of the imperceptible
or seemingly insignificant nature of those accretions, pilots who operate the airplane’s deicing
boots in accordance with manufacturer’s guidance (that advises them to wait until a
recommended thickness of ice accretes) may not activate the deicing boots under these
circumstances. An article written by a Douglas Aircraft Company design engineer (published in
January 1979) indicated that although most pilots are aware of the adverse aerodynamic effects
of large amounts of ice, pilots appear less aware that seemingly insignificant amounts of thin,
rough ice on an airfoil’s leading edge can significantly degrade the airplane’s flight
characteristics. The deicing boot operating procedures now contained in most airplane manuals
contribute to this lack of awareness by advising pilots to wait until a recommended thickness of
ice accretes. During the investigation of this accident, arguments were made that the pilots
caused the accident because they accepted an airspeed 10 knots slower than Comair’s FSM
recommended for holding in icing conditions. However, the Safety Board notes that an EMB120 loaded and configured similar to Comair flight 3272, and operated at 150 knots without any
ice accretions, would have a 36-knot margin between its operating airspeed and the stall speed. This margin would likely appear to be an adequate safety margin to a pilot who did not recognize
that the airplane was accumulating ice or did not believe that enough ice had accumulated to
warrant activation of the deicing boots. The flight handling testing that occurred during the icing
certification process did not identify that control problems that were observed in the accident
airplane’s performance at an airspeed of about 156 knots (only 4 knots below the 160-knot
minimum speed for flight in icing conditions set by the FAA following the Comair accident)
with only a small amount of ice accreted on the deicing boots.
[ANALYSIS] According to the
NCAR report, the area of reduced reflectivity indicated that “the snow-making process was less
efficient there, thus allowing a greater opportunity for liquid cloud to exist.” Postaccident
statements obtained from the other pilots who were operating along the accident airplane’s
flightpath (and passed through the area of low reflectivity) near the time of the accident indicated
that they encountered widely variable conditions. For example, the pilots of Cactus 50 reported
moderate rime icing with the possibility of freezing drizzle, the pilots of NW flight 272
encountered moderate-to-severe rime icing as soon as they leveled off at 4,000 feet msl, and the
pilots of NW flight 483 reported no icing. Comparison of data from the airplanes indicates that the differences in airframe
ice accretion reported by the pilots can be attributed to slight differences in timing, altitude,
location (ground track), airspeed, and icing exposure time (and time within the area of reduced
reflectivity) of the airplanes. Based on weather radar information and pilot statements, the
Safety Board concludes that the weather conditions near the accident site were highly variable
and were conducive to the formation of rime or mixed ice at various altitudes and in various
amounts, rates, and types of accumulation; if SLD icing conditions were present, the droplet
sizes probably did not exceed 400 microns and most likely existed near 4,000 feet msl.
2.4
Aerodynamic Effect of the Ice Accretion on Comair Flight 3272
To help assess the type, amount, and effect of the ice that might have been
accumulated by Comair flight 3272 during its descent, the Safety Board reviewed the available
icing and wind tunnel research data, conducted additional airplane performance
studies/simulations, and requested NASA’s assistance in conducting icing research tunnel (IRT)
143
tests and computational studies.
[CONCLUSIONS > FINDINGS] Because the pilots of Comair flight 3272 were operating the airplane with
the autopilot engaged during a series of descents, right and left turns, power
adjustments, and airspeed reductions, they might not have perceived the
airplane’s gradually deteriorating performance.
11. The accident airplane’s left roll tendency was precipitated by a thin layer of
rough ice that accumulated on the leading edge of the wing during the
airplane’s cruise descent, and was then affected by some or all of the
following factors: the autopilot-commanded left roll, asymmetrical ice selfshedding, aileron deflection effects (localized airflow separations), the
effects of engine/propeller thrust, the asymmetrical power application, and
the disengagement of the autopilot. It is unlikely that the absence of
conductive edge sealer on the left wing leading edge deicing boot segments
was a factor in the airplane’s excessive left roll.
12. Consistent with Comair’s procedures regarding ice protection systems, the
pilots did not activate the leading edge deicing boots during their descent
and approach to the Detroit area, likely because they did not perceive that
the airplane was accreting significant (if any) structural ice.
13. Had the pilots of Comair flight 3272 been aware of the specific airspeed,
configuration, and icing circumstances of the six previous EMB-120 icingrelated events and of the information contained in operational bulletin 120002/96 and revision 43 to the EMB-120 airplane flight manual, it is possible
that they would have operated the airplane more conservatively with regard
to airspeed and flap configuration or activated the deicing boots when they
knew they were in icing conditions.
14. The current operating procedures recommending that pilots wait until ice
accumulates to an observable thickness before activating leading edge
deicing boots results in unnecessary exposure to a significant risk for
turbopropeller-driven airplane flight operations. Based primarily on
concerns about ice bridging, pilots continue to use procedures and practices
that increase the likelihood of (potentially hazardous) degraded airplane
performance resulting from small amounts of rough ice accumulated on the
leading edges.
15. It is possible that ice accretion on unprotected surfaces and intercycle ice
accretions on protected surfaces can significantly and adversely affect the
178
aerodynamic performance of an airplane even when leading edge deicing
boots are activated and operating normally.
16. Current ice detection/protection requirements and application of technology
(particularly deice boots) may not provide adequate protection for a variety
of ice accumulation scenarios (tailplane, supercooled large droplets, thin,
rough ice accumulations, etc.).
17. The guidance provided by Comair in its memos, bulletins, manuals, and
training program did not adequately communicate or emphasize specific
minimum airspeeds for operating the EMB-120 in the flaps-up
configuration, in or out of icing conditions, and thus contributed to the
accident.
18.
[ANALYSIS] Further,
icing of the magnitude described by the pilots of NW flight 272 would have produced strong
visual cues, and it is likely that the pilots would have commented on such a rapid accumulation,
had it occurred. As previously noted, the accident airplane’s CVR did not record any flightcrew
comments about ice accumulation or the need to activate the leading edge deicing boots during
the last 5 minutes of the accident flight; this is consistent with an ice accumulation that was
either not observed by the pilots or that was observed, but considered to be unremarkable.
2.4.1
Possible Factors in Left Roll Tendency
Although the accident airplane’s entire airframe was exposed to roughly
equivalent icing conditions, making it theoretically possible that ice accumulation would be
symmetrical, icing research and wind tunnel tests revealed that ice accumulation (especially ice
accumulated at near freezing temperatures, as occurred in this accident) is rarely totally
symmetrical, either physically or aerodynamically. FDR information showed that, from the time
the airplane leveled at 4,000 feet msl, and then began the left turn, the airplane exhibited an
increasing left roll tendency. The Safety Board identified several factors that might have
contributed in varying degrees to the left roll, asymmetric conditions, and loss of control
observed in this accident. These possible factors included the following:
• asymmetrical ice accumulation effects, possibly caused by ice self-shedding. NASA’s IRT tests indicated that the TAT present at the time of the accident
was likely to result in asymmetrical ice self-shedding. Increased vertical
acceleration values might have exacerbated the ice self-shedding because of
wing bending.
• the aerodynamic effects of aileron deflections on airfoils with an ice
contaminated leading edge (after the left bank was established, the autopilot
commanded right-wing-down aileron deflections—left aileron down, right
aileron up—to resist the steepening left roll) during the left turn. Computational two-dimensional studies and wind tunnel research indicated
that at higher AOAs the downward aileron deflection could initiate a localized
flow separation, which resulted in a decrease in lift on the left wing. The
177 Results from the SLD icing tanker tests suggest that the visual cues for SLD ice accumulations
(unusually extensive ice accreted on the airframe in areas not normally observed to collect ice, accumulation of ice
on the upper surface of the wing aft of the protected area, and on the propeller spinner farther aft than normally
observed) would have been very apparent to the pilots and might have resulted in a comment.
147
localized flow separation occurred at a lower AOA on the downward
deflecting aileron than the upward deflecting aileron, thereby considerably
reducing (or even reversing)178 the rolling moment induced by the wheel
inputs.
[ANALYSIS] Many of the concerns raised about icing in this investigation were previously
identified by the Safety Board as early as its September 1981 study on icing avoidance and
protection. The study raised concerns about the adequacy of the Part 25 appendix C envelope
and icing certification and the difficulties in defining and forecasting icing conditions; as a result
of the study, the Safety Board recommended, in part, that the FAA evaluate individual aircraft
165
performance in icing conditions and establish operational limits, review icing criteria in Part 25
and expand (adjust) the Part 25 appendix C envelope as necessary, and establish standardized
procedures for icing certification. For many years, the FAA did not respond positively to the
Safety Board’s recommendations, indicating that icing was not a significant problem for
airplanes certificated under Part 25 appendix C. However, subsequent icing-related accidents at
Pasco, Washington (in December 1989), and Beckley, West Virginia (in January 1991), revealed
that flight control anomalies could result from tailplane icing (see section 1.18.7) and an icingrelated accident at Cleveland, Ohio (in February 1991), revealed that slightly rough ice
accumulations on the wing upper surface can result in hazardous flight handling
characteristics.188 Further, the October 1994 ATR-72 accident at Roselawn, Indiana,
demonstrated that icing outside the Part 25 appendix C envelope could be a significant problem
for airplanes certificated to operate in icing conditions. After this series of fatal accidents (all of which involved icing in transport
airplanes operated in air carrier service) drew attention to icing-related hazards, the FAA reacted
incrementally to tailplane icing, then rough ice accumulations on the upper wing, and then, later,
to runback icing (SLD). The Safety Board recognizes that following the Comair flight 3272
accident, the FAA began an important icing-related research program with Embraer and the
UIUC. This work has resulted in findings about the effects of thin/rough ice accretions and ice
ridges on boots, with other possible factors (such as intercycle icing and residual ice on boots) as
yet unknown or unresolved. However, had the FAA adequately responded to the Safety Board’s
1981 icing recommendation, the earlier accidents, or the concerns expressed in its own staff’s
draft report on the EMB-120 and conducted a thorough program of icing-related research that
defined a course of action to prevent similar incidents by addressing the certification and
operational issues (autopilot use in icing conditions, no autopilot bank angle exceedence
warning, no stall warning/protection system adjustment for icing conditions, the effects of thin,
rough ice and SLD accretions, etc.), this accident would likely have been avoided.
[ANALYSIS] The Safety Board notes that the failure of the FAA to promptly and systematically
address these certification and operational issues resulted in the pilots of Comair flight 3272
being in a situation in which they lacked sufficient tools (autopilot bank angle warning, adjusted
stall warning/protection system, ice detection system, adequate deice procedures) and
information (airspeed guidance, hazards of thin rough ice accretions, and absence of ice
bridging) to operate safely. The Safety Board concludes that despite the accumulated lessons of
several major accidents and (in the case of the EMB-120) the specific findings of a staff
engineer, the FAA failed to adopt a systematic and proactive (rather than incremental and
reactive) approach to the certification and operational issues of turbopropeller-driven transport
airplane icing, which was causal to this accident.
188 As discussed in section 1.18.1.1, there have been five DC-9 series 10 airplane takeoff accidents
attributed to upper wing ice contamination in the United States since 1968. Although these accidents involved
turbojet-driven airplanes (not turbopropeller-driven airplanes, like the other icing-related incidents/accidents
discussed in this report), the issue of the FAA’s failure to address icing-related operational and certification issues is
pertinent to all airplanes certificated for flight in icing conditions.
166
2.6.2
Icing Certification Requirements
The Safety Board reviewed EMB-120 test data from the original certification of
the airplane for flight in icing conditions (U.S. and Canadian tests) and the subsequent SLD icing
certification tests, which were conducted in 1995 as a result of the ATR-72 accident near
Roselawn, Indiana. The Safety Board found no evidence that the EMB-120 did not satisfy the
tests to which it was subjected; in fact, during these tests, Embraer demonstrated the airplane’s
flight handling qualities under conditions that exceeded the boundaries of the Part 25 appendix C
envelope in terms of LWC. Despite the apparent fulfillment of all icing certification requirements by the
EMB-120, Comair flight 3272 crashed after apparently accreting a thin layer of rough,
“sandpaper-type” ice, in icing conditions that likely fell mostly within the boundaries of Part 25
appendix C, although droplets as large as 400 microns might have been present. Consequently, the Safety Board reviewed the adequacy of the current FAA
requirements for the certification of airplanes for flight in icing conditions. For an airplane to be
certificated for flight in icing conditions, the FAA requires the manufacturer to demonstrate a
limited number of test data points within the Part 25 appendix C envelope. The FAA’s icing
certification requirements are based on fully functioning and operating anti-icing and deicing
systems.
[ANALYSIS] An FAA engineer reviewed these six incidents in a draft report dated January 26, 1996 (see
section 1.18.2.3).
The Safety Board has been unable to obtain information about the specific
disposition of the draft report within the FAA, although the FAA asserted after the accident that
this report did not reflect the official views of the FAA. Nevertheless, the Safety Board notes
that more than 1 year before the accident, at least some members of the FAA certification staff
responsible for handling EMB-120 icing issues were concerned about, and were considering
recommendations on, the following issues: 1) the airplane’s roll behavior with ice accretion, 2)
high drag from ice accretions that are not considered by the flightcrew to warrant activating the
deicing boots, 3) inadequate stall warning in icing conditions, 4) inadequate stall margin with the
airspeed established for use in icing conditions, and 5) problems stemming from the use of the
autopilot in these conditions. The FAA’s official response to the six preaccident EMB-120 icing-related events,
as expressed to the Safety Board by aircraft certification office (ACO) personnel, was that these
incidents shared a common factor—flightcrew failure to activate the leading edge deicing boots. The FAA apparently believed that the EMB-120 was safe to operate in icing conditions as long
as the boots were operated. Hence, the FAA’s primary action regarding EMB-120 icing before the accident
was to approve the Embraer-proposed, CTA-approved revision to the AFM that pilots activate
the boots at the first indication of ice accumulation (revision 43). In doing so, the FAA ACO
apparently did not accept the draft report’s conclusions, which recognized that pilots would not
activate the boots if they did not recognize ice accumulation, that an engaged autopilot masked
the tactile cues of icing, and that under these conditions, the flightcrew also could be deprived of
an adequate stall warning. The Safety Board notes with disappointment that this was the latest in a series of
limited actions taken by the FAA to address the problems of structural icing in transport airplane
certification and operation. Basic knowledge about the aerodynamics of icing (including the
knowledge regarding the hazards of small amounts of surface roughness/ice) has been well
established for the past 50 years (see section 1.18.1), and there is nothing that has been learned in
the most recent, postaccident wind tunnel tests and analyses that could not have been learned
before this Comair accident. Many of the concerns raised about icing in this investigation were previously
identified by the Safety Board as early as its September 1981 study on icing avoidance and
protection.
[ANALYSIS] The Safety Board’s review of data from natural icing flight tests revealed that the
airplane’s handling characteristics were evaluated with ½-inch accretions on protected surfaces
and that the deicing boots’ ability to remove ice accretions of up to ½ inch was assessed. Embraer was not required to demonstrate the EMB-120’s stall characteristics in adverse
operational scenarios, including delayed boot activation, intercycle ice accretion, or residual ice
on boots. As a result of the existing icing certification procedures, the FAA did not account for a
thin ice accumulation (as was identified during this investigation, and which may not be
observed or perceived by pilots to be a threat) that could result in a more hazardous situation
than the 3-inch ram’s horn shape (which is readily recognizable by pilots as a hazard and would
certainly prompt activation of the boots). The Safety Board is concerned that there may be other
unaccounted for ice shapes and/or accretion patterns that could result in potentially hazardous
performance degradation. The Safety Board is also concerned that the current icing certification process is
overly dependent upon pilot performance; the FAA has long based its icing certification policies
and practices on the assumption that pilots will perform their duties without error or
misperception. FAA icing-related publications indicate that if ice formations other than those
considered in the certification process are present, the airplane’s airworthiness may be
compromised. After an airplane is certificated by the FAA for flight in appendix C icing
conditions, it becomes primarily the pilots’ responsibility to ensure that the airplane is operated
in icing conditions for which it was certificated. However, as noted during the investigation of
the ATR-72 accident at Roselawn, during normal flight operations, pilots often cannot tell the
difference between icing conditions that fall within the appendix C envelope and icing conditions
outside the appendix C envelope.191 (For example, a pilot cannot differentiate between 40
micron droplets and 100 micron droplets.) Because pilots often cannot determine whether icing
conditions are consistent with “those considered in the certification process” (i.e., limited points
within the appendix C certification envelope), or not (i.e., SLD icing conditions, or other
potentially hazardous conditions that were not subjected to testing, analysis, or demonstration
during icing certification work), it is virtually inevitable that the airplane will unknowingly be
operated in icing conditions that fall outside the certification envelope, or in which the airplane
had not demonstrated that it could operate safely.
[ANALYSIS] The NASA-Lewis and FAA/UIUC tests indicated that thin, rough ice accretions
located on the leading edge and lower surface of the airfoil primarily resulted in increases in
drag, whereas thin, rough ice accretions located on the leading edge and upper wing surface had
an adverse effect on both lift and drag; this is consistent with information that has been obtained
during NACA/NASA icing research conducted since the late 1930s. Data from research
conducted in the 1940s and 1950s indicate that an airfoil’s performance can be significantly
affected by even a relatively small amount of ice accumulated on the leading edge area, if that
accumulation has a rough, sandpaper-type surface. Consistent with these data, NASA’s drag calculations indicated that the thin,
rough layer of sandpaper-type ice accumulation resulted in significant drag and lift degradation
on the EMB-120 wing section. Further, the thin rough ice accumulation resulted in a decrease in
stall AOA similar to that observed in wind tunnel tests with 3-inch ram’s horn ice shapes on
protected surfaces and frequently demonstrated a more drastic drop off/break at the stall AOA. FAA/UIUC conducted wind tunnel tests using generic shapes to represent the sandpaper-type
roughness with ridges placed on the upper wing surface at 6 percent of the wing chord (farther
aft than the ice ridges observed during NASA’s IRT tests); these tests further demonstrated that
the ridge type of ice accretion resulted in more adverse aerodynamic effect than the 3-inch ram’s
horn ice shapes. As previously noted, NASA’s IRT tests indicated that when an EMB-120 wing is
exposed to conditions similar to those encountered by Comair flight 3272 before the accident,
the airfoil tended to accrete a small ice ridge (or ridges) along the deicing boot tube segment
176 According to NASA-Lewis scientists, some of the frost accretion observed aft of the deicing
boot on the lower wing surface during the icing tunnel tests might have been an artifact of the icing research tunnel
(resulting from the higher turbulence, humidity, and heat transfer characteristics of the tunnel); however, the B.F. Goodrich impingement study and NASA’s LEWICE program (see section 1.16.2) also predicted a sparse, rough ice
accretion aft of the deicing boot on the lower wing surface for some of the tested conditions.
[ANALYSIS] The flight handling testing that occurred during the icing
certification process did not identify that control problems that were observed in the accident
airplane’s performance at an airspeed of about 156 knots (only 4 knots below the 160-knot
minimum speed for flight in icing conditions set by the FAA following the Comair accident)
with only a small amount of ice accreted on the deicing boots. It is possible that if the FAA had
required manufacturers to conduct tests with small amounts of rough-textured ice accreted on the
protected surfaces (as might occur before boot activation and between boot cycles) during icing
certification testing, the absence of an adequate safety margin above the stall speed would have
been identified. Further, the FAA could have ensured pilot awareness of icing and adequate stall
warning by requiring manufacturers to install ice detectors192 and stall warning systems with
reduced AOA thresholds for operations in icing conditions. Based on its concerns that the current icing certification standards did not require
testing for all realistic hazardous ice accretion scenarios, in its 1981 icing-related safety study,
the Safety Board recommended that the FAA review the adequacy of the 1950s-era Part 25
appendix C icing envelope, update the procedures for aircraft icing certification , and oversee the
manufacturers’ evaluations of aircraft performance in various icing conditions. The circumstances of the Comair flight 3272 accident demonstrated again the continuing need for these
FAA actions. The Safety Board considers the information that has been available regarding thin,
rough ice accretions sufficient to have prompted the FAA to require additional testing within
the appendix C envelope to demonstrate the effects of thin, rough ice as part of the icing
certification process. Had the FAA required such additional testing, the resultant information
regarding the stall margin and operational envelope of the EMB-120 might have been used to
define minimum airspeeds for operating the airplane in icing conditions. Therefore, based on its
192 Rosemount ice detectors were first used in military and transport-category airplanes in the early
1970s.
169
review of the history of icing information, the icing-related incident and accident history, the
EMB-120 initial icing certification data, the EMB-120 SLD icing controllability test results, and
the circumstances of this accident, the Safety Board concludes that the icing certification process
has been inadequate because it has not required manufacturers to demonstrate the airplane’s
flight handling and stall characteristics under a sufficiently realistic range of adverse ice
accretion/flight handling conditions. As a result of its investigation of the 1994 Roselawn accident, the Safety Board
issued Safety Recommendations A-96-54 and A-96-56 (currently classified “Open—Acceptable
Response”), which, respectively, stated that the FAA should do the following:
Revise the icing criteria published in 14 CFR Parts 23 and 25, in light of both
recent research into aircraft ice accretion under varying conditions of liquid
water content, drop size distribution, and temperature, and recent
developments in both the design and use of aircraft. Also, expand the
Appendix C icing certification envelope to include freezing drizzle/freezing
rain and mixed water/ice crystal conditions as necessary.
[CONCLUSIONS > FINDINGS] The guidance provided by Comair in its memos, bulletins, manuals, and
training program did not adequately communicate or emphasize specific
minimum airspeeds for operating the EMB-120 in the flaps-up
configuration, in or out of icing conditions, and thus contributed to the
accident.
18. The pilots likely did not recognize the need to abide by special restrictions
on airspeeds that were established for icing conditions because they did not
perceive the significance (or presence) of Comair flight 3272’s ice
accumulation.
19. Whether the pilots perceived ice accumulating on the airplane or not, they
should have recognized that operating in icing conditions at the air traffic
control-assigned airspeed of 150 knots with flaps retracted could result in an
unsafe flight situation; therefore, their acceptance of the 150-knot airspeed
assignment in icing conditions without extending flaps contributed to the
accident.
20. Minimum airspeed information for various flap configurations and phases
and conditions of flight would be helpful to pilots of all passenger-carrying
airplanes.
21. The stall warning system installed in the accident airplane did not provide
an adequate warning to the pilots because ice contamination was present on
the airplane’s airfoils, and the system was not designed to account for
aerodynamic degradation or adjust its warning to compensate for the
reduced stall warning margin caused by the ice.
22. The accident airplane’s autopilot was capable of normal operation and
appeared to be operating normally during the last minutes of the accident
flight, and the autopilot disconnect and warning systems operated in a
manner consistent with their design logic.
23. Had the pilots been flying the airplane manually (without the autopilot
engaged) they likely would have noted the increased right-wing-down
control wheel force needed to maintain the desired left bank, become aware
of the airplane’s altered performance characteristics, and increased their
179
airspeed or otherwise altered their flight situation to avoid the loss of
control.
24. Disengagement of the autopilot during all operations in icing conditions is
necessary to enable pilots to sense the aerodynamic effects of icing and
enhance their ability to retain control of the airplane.
25. If the pilots of Comair flight 3272 had received a ground proximity warning
system, autopilot, or other system-generated cockpit warning when the
airplane first exceeded the autopilot’s maximum bank command limits with
the autopilot activated, they might have been able to avoid the unusual
attitude condition that resulted from the autopilot’s sudden disengagement.
26. Despite the accumulated lessons of several major accidents and (in the case
of the EMB-120) the specific findings of a staff engineer, the Federal
Aviation Administration failed to adopt a systematic and proactive (rather
than incremental and reactive) approach to the certification and operational
issues of turbopropeller-driven transport airplane icing, which was causal to
this accident.
27. The icing certification process has been inadequate because it has not
required manufacturers to demonstrate the airplane’s flight handling and
stall characteristics under a sufficiently realistic range of adverse ice
accretion/flight handling conditions.
28.
[ANALYSIS] The left
wing leading edge deicing boot segments were most recently replaced in July 1996—it strains credibility to presume
that the pilots and mechanics who examined the airplane’s leading edges during preflight and maintenance
inspections between July 1996 and the day of the accident did not observe inconsistently applied conductive edge
sealer. Further, during postaccident interviews, Comair’s maintenance personnel appeared to be very familiar with
the conductive edge sealer application.
148
whether the lack of conductive edge sealer on the upper wing surface at the aft edge of the
leading edge deicing boot had potential to act as a preferred ice accumulation location. The
NASA-Lewis scientists stated that it was unlikely because there were no perceivable tactile
differences (gaps, edges, roughness, etc.) between the leading edge deicing boot and the wing
skin to trigger ice accumulation. More importantly, the NASA-Lewis scientists stated that IRT
tests had demonstrated that ice accretion that far aft on the upper wing surface would be unlikely
to occur in a non-SLD icing environment—it would be more likely to occur in larger SLD
droplet sizes, because of the resultant runback and secondary ice accumulation. The Safety Board concludes that the accident airplane’s left roll tendency was
precipitated by a thin layer of rough ice that accumulated on the leading edge of the wing during
the airplane’s cruise descent and was then affected by some or all of the following factors: the
autopilot-commanded left roll, asymmetrical ice self-shedding, aileron deflection effects
(localized airflow separations), the effects of engine/propeller thrust, the asymmetrical engine
power application, and the disengagement of the autopilot. It is unlikely that the absence of
conductive edge sealer on the left wing leading edge deicing boot segments was a factor in the
airplane’s excessive left roll.
2.5
Flightcrew Actions
2.5.1
Use of Deice/Anti-ice Equipment
The Safety Board attempted to determine whether the airplane’s ice protection
systems were operated during the accident airplane’s descent and approach to Detroit
Metropolitan/Wayne County Airport (DTW). CVR information showed that when the pilots
performed the descent checklist at 1547, they confirmed that the airplane’s “standard seven”
anti-ice systems were activated and activated the windshield heat and the propeller deice
system.180 This was consistent with guidance contained in Comair’s EMB-120 Flight Standards
Manual (FSM), which stated that anti-ice systems should be activated “before flying into known
icing conditions” to prevent ice accumulation on the affected surfaces.
[ANALYSIS] In addition, the Safety Board reviewed wind tunnel test data
obtained during research conducted by the FAA at the University of Illinois at
Urbana/Champaign (UIUC).
The Safety Board’s study of the accident airplane’s aerodynamic performance
indicated that it began to degrade from ice accumulation173 about 4½ to 5 minutes before the
autopilot disengaged, as the airplane descended through 7,000 feet msl; the amount of
degradation increased gradually as the airplane descended to 4,000 feet msl. (See section
1.16.1.2.) Based on this gradual performance degradation, weather radar data that showed light
precipitation intensities, pilot reports of moderate or less ice accretions,174 and the Safety Board
and NCAR weather studies, it appeared likely that Comair flight 3272 encountered icing
conditions that fell within the Part 25 appendix C envelope (see section 1.6.1) and/or the lower
portion of the SLD icing range during its descent to 4,000 feet msl. Thus, the postaccident icing
tunnel tests were performed using LWCs between 0.52 and 0.85 grams per cubic meter and
water droplet sizes between 20 microns and 270 microns. Total air temperatures (TAT) used in
the icing tunnel tests ranged between 26o F and 31o F (-3o C and -0.5o C),175 consistent with the
static air temperature (SAT) values recorded by the FDR during the airplane’s descent from
7,000 to 4,000 feet msl. The exposure time used in the icing tunnel tests was 5 minutes;
additional runs were conducted under some test conditions to determine the effect that deicing
boot activation had on cleaning the leading edge and on subsequent ice accretions. The icing tunnel tests did not result in thick ice accumulation under any test
condition (including SLD droplets); rather, the tests consistently resulted in a thin (0.25 inch
accumulation or less), rough “sandpaper-type” ice coverage over a large portion of the airfoil’s
leading edge deicing boot surface area (and aft of the deicing boot on the lower wing surface in
some test conditions). In addition, in many IRT test conditions, small (½ inch) ice ridges
accreted along the leading edge deicing boot seams. (The effects of these ridges will be
discussed later in this section.) According to NASA and Safety Board IRT test observers, the
thin, rough ice coverages (and ice ridges, where applicable) that accreted on the EMB-120 wing
were somewhat translucent and were often difficult to perceive from the observation window.
[ANALYSIS] The Safety Board notes that FAA Order 7110.10L, “Flight Services,” contains a
definition of “trace” ice accumulations, that states, in part, “A trace of ice is when ice becomes
perceptible….It is not hazardous even though deicing/anti-icing equipment is not utilized unless
encountered for an extended period of time [over 1 hour].” Information obtained during this
investigation, which echoed the results of research conducted in the 1930s and 1940s, indicated
that thin, rough amounts of ice, even in trace amounts can result in hazardous flight conditions. The Safety Board concludes that the suggestion in current FAA publications that “trace” icing is
“not hazardous” can mislead pilots and operators about the adverse effects of thin, rough ice
accretions. Therefore, the Safety Board believes that the FAA should amend the definition of
trace ice contained in FAA Order 7110.10L (and in other FAA documents as applicable) so that
it does not indicate that trace icing is not hazardous. The Safety Board notes that in some icing exposure scenarios, pilots could
become aware of the performance degradation without observing a significant accumulation of
ice on the airplane by observing other cues, such as a decrease in airspeed, excessive pitch trim
usage, a higher-than-normal amount of engine power needed to maintain a stabilized condition,
and/or anomalous rates of climb or descent. However, the Safety Board concludes that because
the pilots of Comair flight 3272 were operating the airplane with the autopilot engaged during a
series of descents, right and left turns, power adjustments, and airspeed reductions, they might
not have perceived the airplane’s gradually deteriorating performance.
146
Further, although it is possible (based on the icing reported by the pilots of NW
flight 272 and the NCAR scientist’s estimation of the likely droplet size distribution in the
clouds) that the accident flight encountered SLD icing177 as it reached 4,000 feet msl, the airplane
was only at that altitude for about 25 seconds before the upset occurred; during most of that 25
seconds, the FDR data showed that the autopilot was countering the increasing left roll tendency
and a sideslip condition was developing. However, even if the accident flight had accumulated
ice at the rapid rate reported by the pilots of NW flight 272 (about ½ inch per minute), the
accident flight could not have accumulated a large amount of ice during the brief period of time
it spent at 4,000 feet before the autopilot disengaged and the loss of control occurred. Further,
icing of the magnitude described by the pilots of NW flight 272 would have produced strong
visual cues, and it is likely that the pilots would have commented on such a rapid accumulation,
had it occurred.
[ANALYSIS] The Safety Board considers it likely that future ice detection/protection systems
will decrease the hazards associated with icing by incorporating ice detection and protection
(automatic activation of deicing boots or anti-icing systems) for individual surfaces, including
the horizontal stabilizers, of all airplanes certificated for flight in icing conditions. However,
because ice accretions and their effects are not yet fully understood, the Safety Board concludes
that current ice detection/protection requirements and application of technology (particularly
deice boots) may not provide adequate protection for a variety of ice accumulation scenarios
(tailplane, SLD, thin, rough ice accumulations, etc.). Therefore, the Safety Board believes that
the FAA should actively pursue research with airframe manufacturers and other industry
personnel to develop effective ice detection/protection systems that will keep critical airplane
surfaces free of ice; then require their installation on newly manufactured and in-service
airplanes certificated for flight in icing conditions.
2.5.2
Airspeed and Flap Configuration Information
Simulator studies conducted during the investigation revealed that the accident
airplane’s decreasing airspeed in icing conditions was critical in the development of the accident
scenario. According to FDR data, the airplane began to exhibit signs of departure from
controlled flight as it decelerated from 155 to 156 knots. Because the pilots accepted an ATC
instruction to slow to 150 knots and maintained a flaps-up configuration, the Safety Board
evaluated the guidance that Comair provided to its EMB-120 pilots on minimum airspeed in the
flaps-up configuration, the Comair flight 3272 flightcrew’s acceptance of this airspeed without
adjusting the airplane’s configuration, and the FAA’s requirements for airplane manufacturers
with regard to minimum airspeeds.
2.5.2.1
Comair’s Airspeed Guidance
During postaccident interviews, some of Comair’s pilot training personnel
indicated that the company’s EMB-120 pilot training emphasized the 160-knot minimum
airspeed for operating in icing conditions, and Comair’s EMB-120 Program Manager told Safety
Board investigators that 170 knots is the only airspeed the company supports for operating with
the landing gear and flaps retracted. Although the Safety Board’s review of the airspeed
guidance contained in Comair’s EMB-120 FSM revealed that it did not contain specific
minimum maneuvering airspeeds for flight in icing conditions and for various airplane
configurations, it did contain general airspeed information in descriptions of normal and nonnormal procedures and maneuvers. For example, the technique outlined in Comair’s FSM for an
instrument landing system (ILS) approach associated the base leg vector position (which was the
accident airplane’s approximate position on the approach before the upset, albeit still about 20
miles from the destination airport) with 170 knots and the flaps 15 configuration.
[CONCLUSIONS > FINDINGS] 3.1
Findings
1.
The pilots were properly qualified and certificated to perform the flight
during which the accident occurred, and each crewmember had received the
training and off-duty time prescribed by the Federal regulations. There was
no evidence of any preexisting medical or behavioral conditions that might
have adversely affected the flightcrew’s performance.
2.
The airplane was certificated, equipped, and dispatched in accordance with
Federal regulations and approved Comair procedures. There was no
evidence of preexisting mechanical malfunction or other failure of the
airplane structure, flight control or other systems, powerplants or propellers
that would have contributed to the accident.
3.
It is likely that the leading edge deicing system was capable of normal
operation during the accident flight.
4.
The Detroit terminal radar approach controllers who were involved with
flight 3272 were properly qualified and certificated. A review of air traffic
control and facility procedures revealed that the controllers followed
applicable air traffic and wake turbulence separation rules, and air traffic
separation was assured during flight 3272’s approach to the runway.
5.
Although the radar ground tracks of Cactus 50 and Comair flight 3272
converged near the accident site, the Safety Board’s review of winds aloft
and wake vortex sink rates indicated that Cactus 50’s wake vortices would
have been above and northeast of Comair flight 3272’s flightpath near the
upset location. Thus, Comair flight 3272 was separated from the vortices
vertically and horizontally, and, therefore, wake turbulence was not a factor
in the accident.
6.
The airplane was aerodynamically clean, with no effective ice accreted,
when it began its descent to the Detroit area.
7.
The weather conditions near the accident site were highly variable and were
conducive to the formation of rime or mixed ice at various altitudes and in
various amounts, rates, and types of accumulation; if supercooled large
droplet icing conditions were present, the droplet sizes probably did not
exceed 400 microns and most likely existed near 4,000 feet mean sea level.
8.
It is likely that Comair flight 3272 gradually accumulated a thin, rough
glaze/mixed ice coverage on the leading edge deicing boot surfaces,
possibly with ice ridge formation on the leading edge upper surface, as the
airplane descended from 7,000 feet mean sea level (msl) to 4,000 feet msl in
177
icing conditions; further, this type of ice accretion might have been
imperceptible to the pilots.
9.
The suggestion in current Federal Aviation Administration publications that
“trace” icing is “not hazardous” can mislead pilots and operators about the
adverse effects of thin, rough ice accretions.
10. Because the pilots of Comair flight 3272 were operating the airplane with
the autopilot engaged during a series of descents, right and left turns, power
adjustments, and airspeed reductions, they might not have perceived the
airplane’s gradually deteriorating performance.
11.
[ANALYSIS] Although the pilots reacted promptly to the autopilot disengagement and applied
control wheel inputs to counter the resultant abrupt left roll, they were not able to regain control
of the airplane because of the airplane’s extreme unusual attitude, the highly dynamic nature of
the subsequent maneuvers, the presence and effect of ice on the wings, and the low altitude at
which the loss of control occurred. The airplane entered an extreme nose-down pitch attitude
from which it did not recover.
2.6
FAA Oversight Issues
The Safety Board’s investigation of this accident raised concerns about the FAA’s
continuing airworthiness oversight of the EMB-120 and the agency’s oversight of icing-related
incidents and accidents involving turbopropeller-driven transport airplanes, the adequacy of
existing FAA regulatory requirements for the certification of transport-category airplanes for
flight into icing conditions (specifically 14 CFR Part 25 appendix C and Section 25.1419), the
FAA’s policies for AFM and air carrier operating manual revisions, and the sharing of
information related to such revisions between the FAA’s certification and flight standards
personnel.
2.6.1
FAA Continuing Airworthiness Oversight Issues
The Safety Board notes that, like the ATR-42 and -72, the EMB-120 exhibited a
history of icing-related upsets/losses of control before being involved in a related fatal accident. At the time of the Comair accident, six icing-related EMB-120 events had been documented, the
first of which occurred in June 1989.187 The Safety Board’s review of these incidents shows that
before the Comair accident, the EMB-120 fleet had experienced repeated instances of roll upsets
associated with ice accumulations that the pilots either did not observe or did not consider
sufficient to prompt activation of the deicing boots.
187 Similarly, before the ATR-72 accident at Roselawn, Indiana, the FAA had been aware of a
number of prior ATR upset events. The FAA had concluded that these incidents were essentially pilot-induced stall
events; however, further investigation revealed that there were more complex airplane controllability issues
involved in the ATR upset events.
164
FAA and Embraer personnel had noted the recurring events, and the FAA
presented a summary of the six events at an FAA/industry meeting (attended by Safety Board
staff) on November 7, 1995. Further, the FAA and Embraer discussed the events with
representatives from Comair and other operators at a meeting on November 15, 1995, and
additional discussion took place during the EMB-120 SLD icing tanker tests in December 1995. An FAA engineer reviewed these six incidents in a draft report dated January 26, 1996 (see
section 1.18.2.3).
The Safety Board has been unable to obtain information about the specific
disposition of the draft report within the FAA, although the FAA asserted after the accident that
this report did not reflect the official views of the FAA.
[ANALYSIS] Comair’s EMB-120 FSM
defined icing conditions as existing “when the OAT [outside air temperature] is +5o C or below
and visible moisture in any form is present (such as clouds, rain, snow, sleet, ice crystals, or fog
with visibility of one mile or less).”
For years, airplane manufacturers have incorporated leading edge deicing boots in
the design of airplanes that are to be certificated for operation in icing conditions; the purpose of
deicing boots is to shed the ice that accumulates on protected surfaces of the airframe. Over the
years, leading edge deicing boots have demonstrated their effectiveness to operators and pilots
by keeping the wing and tail leading edges relatively clear of aerodynamically degrading ice
accumulations, to the point that operators and pilots have become confident that the airplanes can
be flown safely in icing conditions as long as the airplane’s deicing boots are operated (and
functioning) properly. However, based on problems with earlier deicing boot designs (which
180 Although Embraer’s nomenclature identifies the propeller ice protection mechanism as a
deicing system, it functions as an anti-icing system because it is activated before ice accumulates on the airframe.
149
used larger tubes and lower pressures, resulting in slower inflation/deflation rates),
manufacturers, operators, and pilots developed the belief that premature activation of the leading
edge deicing boots could (as cautioned in Comair’s EMB-120 FSM) “result in the ice forming
the shape of an inflated de-ice boot, making further attempts to deice in flight impossible [ice
bridging].” Thus, at the time of the accident, Comair’s (and most other EMB-120 operators’)
guidance indicated that pilots should delay activation of the leading edge deicing boots until they
observed ¼ inch to ½ inch ice accumulation, despite Embraer’s FAA and Centro Tecnico
Aeroespacial of Brazil (CTA) approved EMB-120 Airplane Flight Manual (AFM) revision 43,
which indicated that pilots should activate the leading edge deicing boots at the first sign of ice
accumulation (see discussion later in this section).
The pilots’ activation of the propeller and windshield ice protection systems when
the airplane entered the clouds would indicate that they were aware that the airplane was
operating in icing conditions. If they had activated the leading edge deicing boots, at least some
of the airplane’s degraded performance would have been restored. However, even if the pilots
observed any of the thin, rough ice accretion that likely existed before the loss of control, they
probably would not have activated the deicing boots because Comair’s guidance to its pilots
advised against activating the deicing boots until they observed a thicker ice accumulation.
[ANALYSIS] The Safety Board considers it likely that the pilots would have commented and/or
taken action (such as activating the deicing boots and/or extending the flaps) if they had
perceived an unsafe condition, either as the result of a significant ice accumulation or an unsafe
airspeed assignment for the airplane’s configuration. The Safety Board acknowledges that
increasing the airspeed by some increment (Vref + 5 knots according to Comair’s EMB-120
FSM) when ice accretion is observed is a fairly standard adjustment in the aviation industry, and
Comair’s FSB 96-04 specified a minimum airspeed of 170 knots for holding in icing conditions. However, ATC had not issued holding instructions to the pilots of Comair flight 3272, nor had
ATC indicated that the pilots should expect to receive holding instructions during the approach
to DTW. Therefore, the pilots might not have considered the 170-knot minimum airspeed for
holding in icing conditions. Additionally, as previously discussed, the pilots might not have
recognized that they were operating in icing conditions because it is possible that the accident
airplane accreted a thin, rough layer of glaze ice that was imperceptible to the pilots. Because
there were no comments recorded by the CVR and because the pilots accepted the 150-knot
airspeed assignment without hesitation, comment, or reconfiguration, the Safety Board concludes
that the pilots likely did not recognize the need to abide by special restrictions on airspeeds that
were established for icing conditions because they did not perceive the significance (or presence)
of Comair flight 3272’s ice accumulation. Further, based on the uncertainty regarding minimum
airspeeds exhibited by Comair pilots during postaccident interviews, the Safety Board considers
it likely that under conditions similar to those encountered by the pilots of Comair flight 3272,
other Comair pilots might have accepted the same 150-knot airspeed assignment. Although the Safety Board considers Comair’s airspeed guidance ambiguous and
unclear and acknowledges that the flightcrew might not have perceived that the airplane was
accumulating ice that affected its flight handling characteristics, the Safety Board notes that the
preponderance of the airspeed guidance available to the pilots indicated that EMB-120 operating
airspeeds of 160 or 170 knots were standard for operating without flaps extended under any
(icing or nonicing) conditions. Though these airspeeds were not established minimum airspeeds,
they were the operator’s procedural guidance and the standards to which Comair’s pilots were
trained. The Safety Board considers that any pilot deviations from standard procedures during
flight operations (although not prohibited and not necessarily unsafe) should be accomplished
thoughtfully and with full consideration given to the possible risks involved. In this case,
operating at 150 knots provided the pilots with a reduced safety margin above the airplane’s stall
speed. The reduction in stall margin was especially critical to the accident flight because the
accident airplane had accreted structural ice during its descent, which was having an adverse
effect on the airplane’s performance characteristics.
[ANALYSIS] In this case,
operating at 150 knots provided the pilots with a reduced safety margin above the airplane’s stall
speed. The reduction in stall margin was especially critical to the accident flight because the
accident airplane had accreted structural ice during its descent, which was having an adverse
effect on the airplane’s performance characteristics. The Safety Board notes that the pilots could
have increased the stall margin by extending 15o of flaps and still complied with ATC’s airspeed
assignment. Further, there was no safety or operational reason to avoid extending the flaps.185
185 The Safety Board considered the possibility that the flightcrew avoided extending the flaps
because of guidance to avoid extended operations in icing conditions with flaps extended. However, as previously
discussed, there were numerous indications that the flightcrew was not considering icing as a significant factor in
the airplane’s operation at the time. The Safety Board also considered that the pilots might have believed that they
had already extended the flaps to 15o at the time that they accepted the 150 knot ATC-assigned airspeed. However,
at that time, the airplane was about 20 miles from the destination airport and maintaining an assigned airspeed of
190 knots; thus, the pilots had not received any of the usual (distance and airspeed-related) cues to extend the flaps.
157
The Safety Board considers it critical that pilots take into consideration potential
adverse conditions, and make correspondingly conservative decisions where they are warranted. Although the pilots might not have perceived that the airplane was accumulating any ice, their
activation of the propeller and windshield heat when the airplane entered icing conditions was an
indication that they were aware that they were entering conditions in which ice accumulation
was possible. Based on Comair’s guidance for an ILS approach (which Comair uses during
pilot training) that associates 170 knots with 15o of flaps on the base leg position, and additional
airspeed guidance suggesting airspeeds of 160 to 170 knots for the accident flight’s conditions,
and the pilots’ responsibility to make safe, conservative decisions consistent with flight in icing
conditions, the Safety Board concludes that whether the pilots perceived ice accumulating on the
airplane or not, they should have recognized that operating in icing conditions at the ATCassigned airspeed of 150 knots with flaps retracted could result in an unsafe flight situation;
therefore, their acceptance of the 150-knot airspeed assignment in icing conditions without
extending flaps contributed to the accident.
2.5.2.3
FAA-related Information Regarding Minimum Airspeeds
Because the issue of safe minimum airspeeds is complex and critical to safe flight
operations, in May 1997 the Safety Board issued Safety Recommendation A-97-31, which asked
the FAA to require air carriers to reflect FAA-approved minimum airspeeds for all flap settings
and phases of flight, including flight in icing conditions, in their EMB-120 operating manuals.
[CONCLUSIONS > FINDINGS] The icing certification process has been inadequate because it has not
required manufacturers to demonstrate the airplane’s flight handling and
stall characteristics under a sufficiently realistic range of adverse ice
accretion/flight handling conditions.
28. The work conducted by the Federal Aviation Administration Environmental
Icing National Resource Specialist and the Aviation Rulemaking Advisory
Committee’s icing-related working groups is of crucial importance to the
future safety of icing operations.
29. The potential consequences of operating an airplane in icing conditions
without
first
having
thoroughly
demonstrated
adequate
handling/controllability characteristics in those conditions are sufficiently
severe that they warrant as thorough a certification test program as possible,
including application of revised standards to airplanes currently certificated
for flight in icing conditions.
30. The current Federal Aviation Administration policy allowing air carriers to
elect not to adopt airplane flight manual operational procedures without
clear written justification can result in air carriers using procedures that may
not reflect the safest operating practices.
31. At the time of the Comair flight 3272 accident, pertinent flight standards
personnel (specifically, the principal operations inspector assigned to
Comair) lacked information critical to the continued safe operation of the
EMB-120 fleet and would have been unable to evaluate the need to
180
incorporate airplane flight manual revision 43 or any alternatives proposed
by air carriers.
32. The Federal Aviation Administration’s current EMB-120 flight data
recorder system inspection procedure is inadequate because it allows
existing flight control sensor anomalies to go undetected, and thus
uncorrected.
33. The failure of pilots who encounter in-flight icing to report the information
to the appropriate facility denies other pilots operating in the area the access
to valuable and timely information that could prevent an accident.
34. The Federal Aviation Administration air traffic control system has not
established adequate procedures for the dissemination of icing-related pilot
reports received in the airport terminal environment; these reports should be
incorporated into automatic terminal information service broadcasts so that
all arriving and departing pilots can become aware of icing conditions in the
area.
181 3.2
Probable Cause
The National Transportation Safety Board determines that the probable cause of
this accident was the FAA’s failure to establish adequate aircraft certification standards for flight
in icing conditions, the FAA’s failure to ensure that a Centro Tecnico Aeroespacial/FAAapproved procedure for the accident airplane’s deice system operation was implemented by U.S.-
based air carriers, and the FAA’s failure to require the establishment of adequate minimum
airspeeds for icing conditions, which led to the loss of control when the airplane accumulated a
thin, rough accretion of ice on its lifting surfaces. Contributing to the accident were the flightcrew’s decision to operate in icing
conditions near the lower margin of the operating airspeed envelope (with flaps retracted), and
Comair’s failure to establish and adequately disseminate unambiguous minimum airspeed values
for flap configurations and for flight in icing conditions.
182
[ANALYSIS] Consequently, the Safety Board reviewed the adequacy of the current FAA
requirements for the certification of airplanes for flight in icing conditions. For an airplane to be
certificated for flight in icing conditions, the FAA requires the manufacturer to demonstrate a
limited number of test data points within the Part 25 appendix C envelope. The FAA’s icing
certification requirements are based on fully functioning and operating anti-icing and deicing
systems. Although there is no requirement for manufacturers to consider the effects of delayed
activation of ice protection systems, intercycle or residual ice accumulations, or other variables
that might result in significant aerodynamic effects, Embraer exceeded the minimum FAA
requirements when Embraer tested the EMB-120 with ¾-inch (U.S.) and 1-inch (Canada) ice
accretions/shapes during initial icing certification.189 Certification records indicate that the
EMB-120 successfully exhibited satisfactory flight handling characteristics with 3-inch ram’s
horn ice shapes installed on unprotected surfaces. Further, during the SLD icing controllability
tests, the FAA tested the EMB-120 with quarter-round artificial ice shapes as large as 1 inch
located at the aft edge of the farthest aft inflatable deicing boot segment (to represent ice
accumulated in icing conditions that fall outside the Part 25 appendix C envelope). The airplane
exhibited full lateral controllability and satisfactory stall warning characteristics in this
condition.190
However, Embraer had not demonstrated (nor was the company required by the
certification authorities to demonstrate) the EMB-120’s performance in other ice configurations
that would result from weather conditions within the Part 25 appendix C LWC and droplet size
189 For U.S. (FAA) icing certification, the EMB-120 was tested with ¼ inch, ½ inch, and ¾ inch of
natural ice on protected surfaces, up to 4 inches of natural ice accumulation on unprotected airfoil surfaces, and 3inch ram’s horn artificial ice shapes on unprotected surfaces; except for the ¾-inch natural ice on protected surfaces,
these conditions could be encountered while operating in icing conditions in accordance with procedures outlined in
the EMB-120 AFM. However, for Canadian icing certification, the EMB-120 was tested with artificial ice shapes
representing conditions considered to be outside normal operation with deicing boots activated (1-inch ram’s horn
ice shapes on protected surfaces).
190 Although some control wheel force exceedences were observed, tanker tests identified more
realistic ice shapes; during subsequent tests with the realistic ice shapes, no excessive control wheel forces or other
anomalies were noted.
167
envelope, including realistic ice shapes (or natural ice) representing a thin layer of sandpapertype ice with a small ice ridge (as may have been experienced by Comair flight 3272). As
discussed in section 2.4, postaccident icing and wind tunnel information indicated that with a
small ice ridge along that thin rough surface, the aerodynamic effect on handling and stall
margin/stall warning (reduced stall AOA and rapid decrease in lift) can be worse than any of the
ice shapes that the FAA required for icing certification.
[ANALYSIS] If they had activated the leading edge deicing boots, at least some
of the airplane’s degraded performance would have been restored. However, even if the pilots
observed any of the thin, rough ice accretion that likely existed before the loss of control, they
probably would not have activated the deicing boots because Comair’s guidance to its pilots
advised against activating the deicing boots until they observed a thicker ice accumulation. Therefore, based on CVR information and on the steady degradation of airplane performance
that was clearly uninterrupted by leading edge deicing boot activation, the Safety Board
concludes that, consistent with Comair’s procedures regarding ice protection systems, the pilots
did not activate the leading edge deicing boots during their descent and approach to the Detroit
area, likely because they did not perceive that the airplane was accreting significant (if any)
structural ice. During the postaccident (November 1997) Airplane Deicing Boot Ice Bridging
Workshop, information regarding recent icing tunnel and flight test research into the ice bridging
phenomenon was disseminated and discussed among industry personnel (see section 1.18.4.2).
The recent research revealed that modern turbine-powered airplanes, with their high-pressure,
segmented pneumatic deicing boots, are not at risk for ice bridging.181 However, in April 1996
when Embraer issued (FAA- and CTA-approved) revision 43 to the EMB-120 AFM, the
procedure it recommended—activation of the leading edge deicing boots at the first sign of ice
accretion—was not consistent with traditional industry concerns about ice bridging. According
to the FAA’s EMB-120 Aircraft Certification Program Manager, when the EMB-120 AFM
revision was proposed by Embraer in late 1995, the deicing boot procedural change was very
controversial and generated numerous discussions among FAA and industry personnel. The
FAA’s EMB-120 Aircraft Certification Program Manager stated that the aircraft evaluation
group (AEG) personnel involved in the discussions about the six EMB-120 icing-related events,
the EMB-120 in-flight icing tanker tests, and the deicing boot procedural change were initially
resistant to the deicing boot procedural change because of the perceived potential for ice
bridging.
181 It is important to note that ice bridging may still be a potential hazard for airplanes with older
technology deicing boots that have slower inflation/deflation rates.
150
The Safety Board notes that during the winter of 1995/1996, senior Comair
personnel (and representatives from other EMB-120 operators) were involved in numerous
meetings and discussions regarding the six preaccident icing-related events and that they
subsequently received Embraer’s Operational Bulletin (OB) 120-002/96 and revision 43 to the
EMB-120 AFM, with its controversial deicing boot procedural change.
[ANALYSIS] The studies showed that this effect was most pronounced when a
leading edge ridge of ice was present.
• differences in local airflow over left and right wings because of propeller
thrust—both propellers rotating in a clockwise direction results in
asymmetrical thrust and a left yaw tendency.
• the airplane leveling off at its assigned altitude and slowing to its assigned
airspeed resulted in an increased AOA and movement of the stagnation point
on the leading edge of both wings, which might have exacerbated the effects
listed above.
• left yaw resulting from the asymmetrical engine power application that
occurred less than 3 seconds before the autopilot automatically disengaged. Flight simulations that varied the timing and symmetry of the engine power
application indicated that when power was applied earlier and/or more
symmetrically, the simulator’s bank did not exceed the autopilot’s command
limit, the autopilot did not abruptly disengage, and the upset did not occur. However, flight simulations indicated that the asymmetrical engine power
application observed in the accident FDR data would not have resulted in an
upset if the aerodynamic degradation from ice was not present. (See sections
1.16.1.2 and 2.2.)
• the effects of the autopilot disengagement—when the autopilot (which had
been commanding RWD aileron to resist the left roll) automatically
disengaged, the sudden absence of resistance resulted in a significant increase
in the left roll and roll rate. (See section 2.5.4.)
The Safety Board also considered the possibility that the absence of conductive
edge sealer179 on the left wing leading edge segments (see section 1.12.2) might have contributed
to an asymmetrical ice accretion that would have increased the left roll tendency. Postaccident
examination revealed that the leading edge segments that did not have the conductive edge sealer
applied were as smooth to the touch as any other part of the deicing boot/leading edge; further,
there was no roughness, cracking, splitting, or delamination observed in the area where the
conductive edge sealer should have been. Safety Board staff asked the scientists at NASA-Lewis
178 NASA’s two-dimensional studies indicated that such a reversal is possible at small aileron
deflections and higher AOAs; FDR data indicated that during the last minute before the upset, the aileron deflection
and AOA was frequently changing.
179 It was not possible to determine when and/or how the conductive edge sealer came to be
missing; whether it eroded or wore off during flight, during the accident sequence, as a result of application of
ICEX, or whether the conductive edge sealer was not applied properly during deicing boot installation.
[ANALYSIS] Thus, the Safety Board
believes that the FAA should require the manufacturers and operators of all airplanes that are
certificated to operate in icing conditions to install stall warning/protection systems that provide
186 FAA and NASA wind/icing tunnel data indicate that the NACA 23012 airfoil with a thin layer
of rough ice on the leading edge with a small ice ridge can stall at angles of attack as low as 5o or 6o.
160
a cockpit warning (aural warning and/or stick shaker) before the onset of stall when the airplane
is operating in icing conditions.
2.5.4
Operation of the Autopilot
The Safety Board was unable to positively determine whether the autopilot was
operating properly based on physical evidence (impact damage precluded functional tests).
However, based on FDR data and a review of the autopilot design characteristics, the Safety
Board concludes that the accident airplane’s autopilot was capable of normal operation and
appeared to be operating normally during the last minutes of the accident flight, and the autopilot
disconnect and warning systems operated in a manner consistent with their design logic. The Safety Board evaluated the flightcrew’s use of the autopilot as it affected the
cues presented to the pilots about the impending loss of control and the behavior of the ailerons
as the loss of control developed. The autopilot’s actions during the last seconds before it
disengaged provided some visual cues that could have warned the pilots of the airplane’s
performance degradation. For example, during the 15 seconds before the autopilot disengaged, it
moved the control wheel to command the ailerons to move in a RWD direction, while the flight
instruments (EADIs) and the pilots’ heading selection indicated that the airplane was in a left
bank. Although it would have been possible for the pilots to observe this and deduce that an
anomalous flight condition existed, these visual cues began very gradually and were subtle and
short lived. The control wheel did not move more than 10o, and the roll angle did not exceed 30o
(only slightly greater than the normal autopilot bank limit for the selected left turn), until about 8
seconds before the upset. The deviations from the desired airplane attitude were becoming
noticeable about the time that the pilots were increasing engine power to maintain 150 knots and
continued as the captain directed the first officer’s attention to the airplane’s airspeed (about 5
seconds before the upset). Given this distraction, it is likely that the subtle visual cues that were
available were not adequate to prompt the pilots to take the direct and aggressive action that
would have been necessary to avoid the upset.
[ANALYSIS] When the captain drew the first officer’s attention to the low airspeed indication at
1554:20.8, the airplane’s airspeed had decreased to 147 knots. During the next 2 seconds, the
pilots more aggressively increased the engine power, and a significant torque split occurred; the
torque values peaked at 108 percent on the left engine and 138 percent on the right engine. The
Safety Board considered several possible reasons for the significant torque split, including
uneven throttle movement by the pilots, ice ingestion by the left engine, a misrigged engine, or
an improper engine trim adjustment on the newly installed right engine; however, it was not
possible to positively determine the cause of the torque split. Postaccident simulations indicated
that this torque split had a significant yaw-producing effect at a critical time in the upset event,
exacerbating the airplane’s excessive left roll tendency (see section 2.4.1). The airplane’s
airspeed decreased further to 146 knots, the left roll angle increased beyond the autopilot’s 45o
limit, and (at 1554:24.1) the autopilot disconnect warning began to sound. One second later, the
stick shaker activated. The sudden disengagement of the autopilot (at 1554:24.125) greatly
accelerated the left rolling moment that had been developing, suddenly putting the airplane in an
unusual attitude. Although the pilots were likely surprised by the upset event, interpretation of
the FDR data indicated that the pilots responded with control wheel inputs to counter the left roll
within 1 second of the autopilot disengagement and continued to apply control inputs in an
apparent attempt to regain control of the airplane until the FDR recording ceased.
2.3
Meteorological Factors
Although Comair flight 3272 was operating in winter weather conditions
throughout its flight from the Cincinnati area to Detroit, CVR and weather information indicated
that the airplane was operating above the cloud tops at its cruise altitude of 21,000 feet msl. Further, the temperatures at the altitudes flown during the en route phase of the flight were too
cold to be conducive to airframe ice accretion, and examination of the FDR data did not reflect
degraded airplane performance until later in the airplane’s descent (see section 1.16.1.1).
Therefore, the Safety Board concludes that the airplane was aerodynamically clean, with no
effective ice accreted, when it began its descent to the Detroit area. A study conducted by the National Center for Atmospheric Research (NCAR)
indicated that there was strong evidence for the existence of icing conditions in the clouds along
the accident airplane’s descent path below 11,000 feet msl.